Hi Luis,

The key (if you’ll forgive the pun) is the behaviour of the “det” function, 
which is (at this writing) defined in Basic/MatrixOps/matrixops.pd as:

sub det {
  my($x) = shift;
  my($opt) = shift;
  $opt = {} unless defined($opt);
  my($lu,$perm,$par);
  if(exists ($opt->{lu}) and (ref $opt->{lu} eq 'ARRAY')) {
    ($lu,$perm,$par) =  @{$opt->{lu}};
  } else {
    ($lu,$perm,$par) = lu_decomp($x);
    $opt->{lu} = [$lu,$perm,$par]
      if(exists($opt->{lu}));
  }
  ( (defined $lu) ? $lu->diagonal(0,1)->prodover * $par : 0 );
}

You will see the last line returns a Perl scalar 0 if the $lu came back 
undefined, otherwise it will do a prodover on the diagonal, which returns a PDL 
ndarray. Here is my slightly-modified version of your script:

use strict; use warnings; use feature 'say';
use PDL;
my $m1=pdl[[-2,-2,-2], [-1,-2,-2], [0,-2,-2]];
my $m2=pdl[[-2,-2,-2],[-1,-1,-2],[0,0,-2]];
for($m1, $m2){
    my $det = $_->det(my $opt={lu=>undef});
    say "Got det=$det (Ref=@{[ref $det]}) lu=(@{$opt->{lu}}) for Matrix=$_";
}

I will look in a while at why the lu_decomp is behaving differently for the two 
matrices. It is almost certainly a bug.

Best regards,
Ed

From: Luis Mochan<mailto:moc...@icf.unam.mx>
Sent: 11 February 2023 04:20
To: perldl<mailto:pdl-general@lists.sourceforge.net>; 
perldl<mailto:pdl-de...@lists.sourceforge.net>
Subject: [Pdl-devel] inconsistent determinant

I found a curious situation when running pdl, as illustrated in the
following example.  I build two singular 3x3 matrices and calculate
their determinant, which is zero. However, in one case I get a PDL
scalar and in the other I get a Perl scalar. I don't know why the
different behaviour:

# Program
use v5.36;
use PDL;
my $m1=pdl[[-2,-2,-2], [-1,-2,-2], [0,-2,-2]];
my $m2=pdl[[-2,-2,-2],[-1,-1,-2],[0,0,-2]];
for($m1, $m2){
    say "Matrix=$_", "Det=", $_->det, "\nRef=", ref $_->det;
}


#Results:
Matrix=
[
 [-2 -2 -2]
 [-1 -2 -2]
 [ 0 -2 -2]
]
Det=0
Ref=PDL
Matrix=
[
 [-2 -2 -2]
 [-1 -1 -2]
 [ 0  0 -2]
]
Det=0
Ref=

I'm currently running PDL v2.081 under Perl 5.36 in Linux/testing.

Best regards,
Luis



--

                                                                  o
W. Luis Mochán,                      | tel:(52)(777)329-1734     /<(*)
Instituto de Ciencias Físicas, UNAM  | fax:(52)(777)317-5388     `>/   /\
Av. Universidad s/n CP 62210         |                           (*)/\/  \
Cuernavaca, Morelos, México          | moc...@fis.unam.mx   /\_/\__/
GPG: 791EB9EB, C949 3F81 6D9B 1191 9A16  C2DF 5F0A C52B 791E B9EB


_______________________________________________
pdl-devel mailing list
pdl-de...@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/pdl-devel

_______________________________________________
pdl-general mailing list
pdl-general@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/pdl-general

Reply via email to